High Efficiency Multicrystalline Silicon Solar Cells: Potential of n-type Doping

Transcription

1 (c) 215 IEEE. DOI: 1.119/JPHOTOV Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material 1 High Efficiency Multicrystalline Silicon Solar Cells: Potential of n-type Doping Florian Schindler, Jonas Schön, Bernhard Michl, Stephan Riepe, Patricia Krenckel, Jan Benick, Frank Feldmann, Martin Hermle, Stefan W. Glunz, Wilhelm Warta, and Martin C. Schubert Abstract In this work, we demonstrate the potential of multicrystalline (mc) n-type silicon for the fabrication of highly efficient mc-si solar cells. High quality mc n-type silicon wafers are obtained from a research ingot crystallized in a high-purity crucible, using high-purity granular silicon as seed layer in the crucible bottom and high-purity polysilicon feedstock for the block. A mc p-type silicon block crystallized under identical conditions (same seed and feedstock, crucible system, and temperature profiles) serves as reference and enables measurements of the interstitial iron and chromium concentrations by metastable defect imaging. In combination with 2-D simulations for in-diffusion and precipitation of chromium, the limitation of n-type high performance multicrystalline silicon (HPM-Si) by these metals is assessed after different solar cell processing steps. Material-related efficiency losses are assessed by an Efficiency limiting bulk recombination analysis (ELBA), which combines injection-dependent photoluminescence imaging of minority charge carrier diffusion length with PC1D cell simulations. Finally, based on this material, boron-diffused front-junction mc n-type silicon solar cells with a full area passivated rear contact (TOPCon) are fabricated. The record cell features an efficiency of 19.6%, which is the highest efficiency reported for an mc n-type silicon solar cell. Index Terms High-performance multicrystalline silicon (HPM-Si), n-type, silicon solar cells, TOPCon I. INTRODUCTION OR the fabrication of high-efficiency silicon solar cells, high quality bulk material with low recombination activity is inevitable. Regarding the type of doping, n-type silicon F Manuscript received May 29, 215; revised July 24, 215; accepted August 4, 215. This work was funded by the German Federal Ministry for Economic Affairs and Energy within the research project THESSO under contract number as well as ForTeS under contract number F. Schindler is with the Fraunhofer Institut für Solare Energiesysteme, 7911 Freiburg, Germany and also with Freiburger Materialforschungszentrum, Albert-Ludwigs-Universität Freiburg, 7914 Freiburg, Germany ( florian.schindler@ise.fraunhofer.de). J. Schön, B. Michl, S. Riepe, P. Krenckel, J. Benick, F. Feldmann, M. Hermle, S.W. Glunz, W. Warta, and M.C. Schubert are with Fraunhofer Institut für Solare Energiesysteme, 7911 Freiburg 7911, Germany ( jonas.schoen@ise.fraunhofer.de; bernhard.michl@ise.fraunhofer.de; stephan.riepe@ise.fraunhofer.de; patricia.krenckel@ise.fraunhofer.de; jan.benick@ise.fraunhofer.de; frank.feldmann@ise.fraunhofer.de; martin.hermle@ise.fraunhofer.de; stefan.glunz@ise.fraunhofer.de; wilhelm.warta@ise.fraunhofer.de; martin.schubert@ise.fraunhofer.de). features advantages like its higher tolerance towards many metal impurities such as interstitial iron due to their lower capture cross sections for minority carriers [1], and the absence of the boron-oxygen-related degradation [2]. On the other hand, minority electron mobility in p-type silicon exceeds minority hole mobility in n-type silicon by a factor of ~2-3 (depending on the resistivity). Thus, in order to reach similar minority carrier diffusion lengths in both materials, the excess carrier lifetime in n-type silicon has to exceed excess carrier lifetime in p-type silicon by this factor to compensate for the lower minority carrier mobility. While this is easy to achieve in monocrystalline silicon, in multicrystalline n-type silicon the advantage of smaller capture cross sections of several dissolved metal impurities is assumed to be less pronounced, as a significant fraction of charge carriers in mc silicon recombines via decorated structural crystal defects, such as dislocation clusters and grain boundaries. Recombination at such crystal defects plays an important role in mc n-type silicon, as for example shown in [3, 4]. In addition, if impurities are present in the form of precipitates, their recombination activity is found to be comparable in mc silicon of both doping types. Metal precipitates will be charged by majority charge carriers, thus attracting minority charge carriers and thereby being an effective recombination center in both mc p- and n-type silicon [5, 6]. Still, high minority carrier lifetimes have been reported for mc n-type Si [7]. Further, for a crystallization in standard industrial quality crucibles, the smaller impact of certain transition metals on minority carrier lifetime in n-type silicon can be of notable advantage concerning recombination via dissolved impurities within the grains, and thus lead to a superior efficiency potential of mc n-type silicon compared with mc p-type silicon [8]. Additionally, progress in crystallization techniques has led to a significant increase in mc silicon material quality, resulting in a new efficiency world record on mc p-type Si [9, 1]. Concluding that the efficiency potential of mc n-type silicon is mainly limited by structural crystal defects decorated with impurities [8], the use of high performance multicrystalline silicon (HPM-Si) with an optimized crystal structure featuring a very low density of dislocation clusters should further decrease recombination losses in mc n-type silicon. Additionally, the use of a high-purity crucible for directional solidification reduces the impact of impurities from the crucible system, which is the main contamination source [11, 12].

2 (c) 215 IEEE. DOI: 1.119/JPHOTOV Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material 2 In this work, we have investigated systematically the recombination properties and efficiency potential of n-type HPM-Si and demonstrate the applicability of a high-efficiency n-type solar cell concept with a full area Tunnel Oxide Passivated rear Contact (TOPCon) [13] on multicrystalline substrate, which is able to exploit the high electrical material quality of n-type HPM-Si. After introducing the investigated material and the applied characterization methods in section II, we focus on the material-related limitations in n-type HPM-Si and review the impact of the doping type on the material quality of mc silicon in section III. Finally, the efficiency potential of the material under investigation is assessed by an Efficiency limiting bulk recombination analysis (ELBA) [14], and TOPCon solar cell results on n-type HPM-Si are presented and discussed in detail in section IV. II. MATERIAL AND METHODS A. Crystallization, lifetime sample and solar cell processing For our investigations, a G1 size n-type HPM-Si ingot (width 22 cm; height 12.8 cm; weight ~14 kg) with a homogeneous distribution of small grains and a low density of dislocations was produced from high-purity feedstock by seed assisted growth in a laboratory crystallization furnace adapted for the crystallization of G1 (~14 kg of Si) and G2 size (~75 kg of Si) silicon ingots at Fraunhofer ISE. In order to obtain material with a very low impurity concentration, a high-purity crucible and coating, high-purity granular silicon as seed layer in the crucible bottom, and high-purity polysilicon feedstock for the main volume were chosen. A p-type HPM-Si ingot, crystallized under identical conditions (same seed and feedstock, crucible system, and temperature profiles; differing only in the type of doping) served as reference. The resistivity of the p-type block is in the range of Ωcm, whereas the n-type block features a stronger variation along the block height from.5 to 1.3 Ωcm due to the smaller segregation coefficient of phosphorus. As we expect a similar impurity concentration in the p-type sister block due to the identical crystallization conditions, metastable defect imaging of interstitial iron and chromium at wafers from the p-type block should give access to their concentrations also in the n-type block cm² bricks were cut symmetrically from the center of the blocks (cf. Fig. 1) to avoid an impact of the edge region of low lifetime affected by solid-state in-diffusion of impurities from the crucible on the investigated wafers (gray shaded area in Fig. 1). Due to size-restrictions of certain processing steps, cm² wafers were cut from the original cm² wafers, and for cell processing, seven 2 2 cm² cells were fabricated on round 4-inch wafers cut symmetrically from the cm² wafers at Fraunhofer ISE s cleanroom. The investigated samples and fabricated solar cells are sketched in Fig. 1. Lifetime samples were fabricated from wafers from the upper third of the blocks corresponding to a net doping concentration of approximately cm -3 (p-type) and (7-8) 1 15 cm -3 (n-type). These wafers were subjected to a variety of solar cell process steps suited for the production of high-efficiency solar cells as indicated in Fig. 2a. Wafers without high-temperature treatment from the initial group served as references. 22 cm 15.6 cm 12.5 cm fabricated solar cells 2 cm investigated wafers central brick crystallized block Fig. 1. Sketch of a crystallized block (top view). The investigated wafers are depicted by the red square, 2 2 cm² solar cells were fabricated on round wafers from the central part. a) Lifetime sample processing b) Cell processing B+P+ TOPCon solar cells Saw Damage Etch Mask Rearside B-diffusion, 89 C, 1h Mask Etch Mask Frontside P-diffusion, 8 C, 1h Mask Etch 8 C, 1h Emitter Etch Surface Passivation Fig. 2. a) Processing of lifetime samples. Cleaning steps are not listed, hightemperature steps are highlighted in red. The surface passivation is Al 2O 3 for the p-type wafers and SiN x for the n-type wafers. b) Processing of mc n-type TOPCon solar cells. The choice of process steps includes a sequence suited for the fabrication of high-efficiency TOPCon solar cells ( B+P+ ). Note that the standard process sequence for TOPCon solar cells does not necessarily include a phosphorus diffusion, which was included in our investigations to apply an additional gettering step. However, as will be discussed later, for the fabrication of high-efficiency TOPCon solar cells on mc n-type substrate, this step is not urgently required. The treatment with all high-temperature steps also included in the cell processing ensures that the material quality of lifetime samples corresponds to that of the final cell and thus allows for a thorough analysis of the material-related efficiency losses. Furthermore, process sequences including solely a boron diffusion at 89 C or a phosphorus diffusion at 8 C were applied to study their impact on the material quality separately. In order to obtain an injection-independent surface passivation of the lifetime samples, p-type wafers were passivated with Al 2 O 3 -films and n-type wafers with SiN x - films. crucible Isotexture Frontside Mask Rearside B-diff.,89 C, 1h Mask Etch TOPCon, 8 C, 1h Passivation Frontside Metallization

3 (c) 215 IEEE. DOI: 1.119/JPHOTOV Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material 3 Seven 2 2 cm² TOPCon solar cells were fabricated on several parallel n-type wafers of medium block height as sketched in Fig. 1, according to the cell process sequence shown in Fig. 2b. The key feature of the TOPCon cell is the ultra-thin wet chemical oxide layer (~14 Å) grown at the rear side in combination with a 15 nm thick phosphorus-doped Si layer deposited on top of the tunnel oxide. This TOPCon layer provides a highly efficient passivated rear contact featuring an excellent interface passivation, an efficiently doped layer for maintaining the quasi-fermi level separation (for high V oc ), as well as an efficient majority carrier transport (for high fill factors) by tunneling through the ultra-thin oxide layer [15]. Lifetime samples were also fabricated with the same process sequence except for the metallization from parallel wafers. For this TOPCon cell batch on multicrystalline silicon, we applied the standard process sequence without phosphorus gettering diffusion in order to reduce process complexity to a minimum, although an additional P-gettering would result in a better material quality, as will be shown in section III. The main difference compared with the TOPCon process on monocrystalline silicon is the surface texturing. For our mc samples, we applied a double-sided acidic texture prior to a planarization of the rear side by chemical polishing, where the tunnel oxide passivated contact will be located. The boron diffusion for the front emitter formation at 89 C differs from the advanced B-diffusion featuring a drive-in oxidation, which can be applied for monocrystalline silicon. In our work, the high-temperature drive-in oxidation was excluded, as the mc material quality would suffer from this step. After the boron diffusion, the TOPCon layer is deposited at the rear side followed by an annealing step at 8 C. Finally, after deposition of an Al 2 O 3 -SiN x -stack serving as passivation and anti-reflection layer on the front side, front and rear side metal contacts were formed. B. Characterization Methods The material quality of the lifetime samples was investigated by photoluminescence imaging (PLI) calibrated by harmonically modulated photoluminescence [16] in order to obtain images of the minority charge carrier lifetime and diffusion length. While the minority charge carrier lifetime delivers a more direct access to the recombination properties in the samples, the minority charge carrier diffusion length is preferred for a comparison of n- and p-type material quality, as it accounts for the large difference in minority electron and hole mobility. On p-type samples, metastable defect imaging by calibrated PLI allows for spatially resolved measurements of the concentrations of interstitial iron (Fe i ) [17] and chromium (Cr i ) [18]. The SRH-parameters for Fe i are taken from [19], for FeB from [2], and for Cr i and CrB from [21]. Due to the identical block crystallizations, this also gives access to their concentrations in the n-type block. Further insight into the recombination properties of HPM-Si is gained by injection-dependent calibrated PLI. This also enables a prediction of the solar cell efficiency potential by an Efficiency limiting bulk recombination analysis (ELBA) [14] (with the modifications described in [22]): Efficiency maps can be calculated by combining injection-dependent images of bulk recombination with PC1D cell simulations. Based on the spatially resolved efficiency potential, specific bulk-related defects limiting the solar cell efficiency can be assessed. We performed ELBA analyses for n-type HPM-Si lifetime samples from the TOPCon process sequence with or without the P-diffusion. The parameters for the PC1D model were taken from recombination current prefactor (J ) measurements of the TOPCon rear side (J,rear = 7 fa/cm²) and the isotextured boron diffused front side (J e = 11 fa/cm²), reflection measurements of an isotextured front side with identical anti-reflection coating and base resistivity measurements of the mc-si material (n = (7-8) 1 15 cm -3 ). Further, a cell thickness of 16 µm and a series resistance of.55 Ωcm² were assumed in the PC1D simulations. Thus, the cell limit without bulk recombination losses would be 2.9%. The difference to the much higher efficiency potential of TOPCon reported in [13] results from the higher emitter recombination as no drive-in oxidation can be applied to mc- Si samples and the higher optical reflectance of the acidic isotexture compared with an alkaline pyramidal texture. For an estimation of the efficiency gain by improvements of the front surface reflectance, in a second simulation we applied the front surface reflection of a random pyramid surface, while keeping all other parameters unchanged. In this case, the cell limit without bulk recombination losses is significantly increased to 22.4%. Although on mc silicon a random pyramid structure cannot be achieved, surfaces with similarly low reflectance such as a honeycomb structure could be applied. III. MATERIAL QUALITY OF HPM-SI A. Lifetime and diffusion length In a first step, we compare minority carrier lifetime and diffusion length in p- and n-type HPM-Si after the different high-temperature steps from Fig. 2a. Fig. 3 illustrates square root harmonic mean (lifetime) and harmonic mean (diffusion length) values averaged across a representative central quarter stripe ( cm²) of the lifetime samples, obtained at constant illumination of.5 suns, which is estimated to correspond to an injection level close to MPP conditions in the solar cell. Minority carrier diffusion length images after the corresponding process steps are shown in Fig. 3c. The n-type HPM-Si samples feature significantly higher minority carrier lifetimes after all processing steps (Fig. 3a). The remarkably high lifetime in the initial state (lifetime within the grains up to 8 µs, square root harmonic mean across the wafer > 35 µs, exceeding the initial lifetime in p-type by more than a factor of five) reflects the larger tolerance of mc n-type silicon towards typical metal impurities present in the material after crystallization compared with mc p-type silicon. These large differences in lifetime are partly compensated by the larger minority carrier mobility in p-type, thus leading to smaller differences in terms of diffusion lengths. Still, in the initial state and after a boron diffusion, also the average diffusion lengths are significantly larger in the n-type samples than in the corresponding p-type samples.

4 (c) 215 IEEE. DOI: 1.119/JPHOTOV Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material 4 a) τ sqrt harm (µs) b) L diff,harm (µm) c) p-type B+P+ B+P+ p-type l B+P+ L diff (µm) 1 n-type B+P+ B+P+ n-type Fig. 3. Minority carrier lifetime (a) (square root harmonic mean) and diffusion length (b) (harmonic mean) at an illumination of.5 suns, averaged across a representative central quarter stripe ( cm²) of p- and n-type HPM- Si lifetime samples after the processing steps of Fig. 2a. The corresponding minority carrier diffusion length images of the central stripe of the wafers are shown below the graph (c). The drop in material quality after boron diffusion compared with the initial state is attributed to the high temperature (89 C) of the applied B-diffusion and its lower gettering efficiency compared to a P-diffusion, which strongly depends on the process parameters [23]. We also observed this effect on standard p- and n-type mc silicon and discussed it in detail in a recent publication [8]. A look at the spatially resolved diffusion length (cf. Fig. 3c) reveals that the drop in the average diffusion length is caused by a strong increase in recombination activity of structural crystal defects (grain boundaries, dislocation clusters), whereas the diffusion length within good n-type grains even improves. This effect will be discussed in more detail in section III.B. It should be noted that despite the decrease in material quality, average diffusion lengths across the whole wafer are still above 5 µm in n-type HPM-Si, which is of importance for the applicability of the standard TOPCon solar cell process to this material, as we will discuss in section IV. After processes including a phosphorus diffusion, minority carrier diffusion lengths strongly increase in the p-type samples, which is an indication to a limitation by a getterable l B+P+ metal impurity such as interstitial iron or chromium. The fact that the phosphorus gettering barely improves the material quality of the n-type samples compared with the initial state additionally points at a limitation by a metal impurity with a larger capture cross section for electrons than for holes, which supports our assumption that interstitial iron could be a candidate for a lifetime limiting defect in p-type HPM-Si before gettering. The limitation of p- and n-type HPM-Si by iron and chromium will be discussed in the next subsection III.B. After processes including a P-diffusion, both p- and n-type samples feature average diffusion lengths above 6 µm, exceeding 1 µm in good grains (cf. Fig. 3b and Fig. 3c), which highlights the excellent electrical quality of the investigated material. B. Limitations of minority carrier diffusion length As mentioned in the previous subsection, the limitation of minority carrier lifetime in p-type HPM-Si is due to a getterable metal impurity with a large capture cross section for electrons, which makes interstitial iron a probable candidate [19]. Additionally, interstitial chromium is a getterable impurity [24] which is harmful in both p- and n-type silicon [25] and contributes significantly to recombination in asgrown grains of standard mc n-type silicon [26]. Therefore, it should also be considered as possible candidate for a dominant defect in p- and n-type HPM-Si in the initial state. To investigate the limitations by Fe i and Cr i in detail, their concentrations were measured by means of metastable defect imaging on the p-type samples in the initial state, after boron diffusion and after phosphorus diffusion. As explained in section II, we assume the same concentrations as measured in the p-type samples also in the n-type wafers. This allows for assessing the fraction of recombination due to interstitial iron ( FoR Fei ) compared to the total recombination in the samples of both doping types by comparing the SRH-lifetime calculated from the average concentration of Fe i with the measured effective lifetime (FoR Fei = τ eff /τ SRH,Fei ). In the same way, the fraction of recombination due to interstitial chromium ( FoR Cri ) in the n-type samples and chromiumboron-pairs ( FoR CrB ) in the p-type samples can be assessed. Note that in thermal equilibrium at room temperature, the majority of dissolved chromium atoms are present as chromium-boron-pairs in the p-type samples, which is why we focus on the limitation due to CrB-pairs rather than Cr i here. In the following, the limitations due to Fe i and CrB or Cr i in p- and n-type HPM-Si are discussed before and after processing. Before high-temperature processing In the initial state, we measure an average Fe i concentration of cm -3. Chromium imaging reveals a CrB/Cr i concentration below the detection limit, which is in the range of cm -3 for the investigated samples. This is in agreement with the concentration of interstitial chromium of cm -3 obtained from 2-D simulations for the indiffusion of chromium from the high-purity crucible system and precipitation during crystallization. The simulation model is explained in detail in [26].

5 (c) 215 IEEE. DOI: 1.119/JPHOTOV Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material p-type initial state FoRFei, avg. = 35% n-type initial state FoRCrB, avg. = 7% FoRFei, avg. = 1% FoRCri, avg. = 7% FoRFei (%) FoRCrB (%) FoRFei (%) FoRCri (%) a) b) c) d) Fig. 4. Fraction of recombination due to interstitial iron (a) and chromiumboron pairs (b) in the initial state of p-type HPM-Si, and fraction of recombination due to interstitial iron (c) and chromium (d) in the initial state of n-type HPM-Si at a generation rate corresponding to.5 suns. Note the different scale of (a). With the measured Fei-concentration, the simulated Criconcentration, and the measured lifetime images, we can calculate images of the fraction of recombination due to Fei and CrB/Cri at an illumination intensity of.5 suns, which are shown in Fig. 4 for the p-type wafer (a, b) and for the n-type wafer (c, d). In the p-type sample, a large part of recombination within grains can be attributed to Fei (FoRFei 4%), whereas the impact of chromium is smaller (FoRCrB 1%). In contrast, due to its significantly smaller capture cross section for holes, an interstitial iron concentration of cm-3 does not contribute to recombination in the n-type sample (τsrh,fei 4 ms). However, the relative impact of Cri in good grains of the n-type sample (FoRCri 1%) is comparable with the relative impact of CrB in grains of the p-type sample. Despite the slightly lower capture cross section for holes of Cri compared with the capture cross section for electrons of CrB, the higher lifetime level in the n-type sample leads to a similar relative limitation by chromium, which highlights the important role of chromium for both doping types. Thus, we can conclude that prior to high-temperature processing, about 5% of recombination in grains of p-type HPM-Si is attributed to Fei (4%) and CrB (1%). A similar contribution to recombination in grains of n-type HPM-Si is found for Cri (1%), whereas recombination at interstitial iron is negligible. After high-temperature treatment Cri is gettered effectively both by the boron or the phosphorus diffusion, such that its concentration is significantly lower than the detection limit (in the range of 19 cm-3) and does not contribute to recombination neither in the p-type nor in the n-type HPM-Si samples after these processes. The average Fei concentration slightly increases after B-diffusion to 4 11 cm-3, which could be attributed to a potential cross-contamination from the furnace tube at elevated temperatures, as suggested in [23]. Also, the higher temperatures of the boron diffusion could lead to a dissolution of precipitates in the vicinity of structural crystal defects, which would lead to a lifetime degradation in these areas. As we observe a homogeneous increase of the Fei concentration 5 across the whole wafer, the first explanation could hold for Fei, whereas the dissolution of other metal precipitates could be responsible for an increased recombination in areas of structural crystal defects. While the Fei concentration of 4 11 cm-3 is responsible for about 5% of recombination within the grains of the p-type sample, it still does not contribute significantly to recombination in n-type HPM-Si. After B-diffusion, the n-type sample even features an increased lifetime within the grains compared to the initial state (cf. Fig. 3c), which could be partly attributed to an effective gettering of Cri by the applied B-diffusion. In contrast, in the p-type sample the reduced concentration of CrB is compensated by the slight increase in the Feiconcentration, thus leading to very similar lifetimes within the grains before and after B-diffusion (cf. Fig. 3c). In both samples, the decrease in average lifetime after B-diffusion can be explained by a strongly increased recombination at structural crystal defects, such as grain boundaries and dislocation (cf. Fig. 3c). As mentioned before, a dissolution and redistribution of previously large metal precipitates to a higher density of smaller metal precipitates could be a possible explanation and highlights the limitation of mc n-type silicon by precipitates and structural crystal defects rather than by dissolved impurities. At this point we can conclude that the asymmetric capture cross sections of Fei for electrons and holes are a major reason for the superior material quality of n-type compared to p-type HPM-Si after processes without Pdiffusion (cf. Fig. 3). After processes including a P-diffusion, the concentration of interstitial iron is considerably below 11 cm-3 and thus of minor importance for recombination even in p-type HPM-Si, which explains the strong increase in minority carrier diffusion length in the p-type wafers after P-gettering. However, even a very low concentration of Fei in this range could still have a slight impact on the stronger injectiondependence of minority carrier lifetime, which we observe in the p-type samples, compared with the rather flat injectiondependence measured on n-type samples [27]. These results and a detailed discussion are included in a separate publication [27]. Thus we can conclude that Fei is a dominant defect in p-type HPM-Si before processing and after B-diffusion, whereas after processes including a P-diffusion it is only of minor impact for recombination. This explains the large increase in diffusion lengths in the p-type material after Pdiffusion. The n-type HPM-Si starts at a very high diffusion length level before processing and is not affected by recombination at Fei, whereas Cri can slightly contribute to recombination in the initial state. However, Cri is effectively gettered by a B-diffusion or a P-diffusion, thus becoming of minor importance for recombination after processing. Especially after the boron diffusion, it is mainly recombination active structural crystal defects that limit the material quality of n-type HPM-Si. After processes including a phosphorus diffusion, p- and n-type HPM-Si end up at a very similar level of high diffusion lengths. A detailed investigation on further metal limitations in

6 (c) 215 IEEE. DOI: 1.119/JPHOTOV Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material multicrystalline n-type silicon has been published recently [26]. In that work, we investigated a standard mc n-type blocks of size G2 (width 4 cm; height 22.5 cm), crystallized from high purity feedstock in standard crucibles. We observed that recombination at FeSi2-precipitates plays a dominant role in the edge region of mc n-type Si, whereas Cu3Si- and NiSi2precipitates may significantly contribute to recombination in the block center. For the HPM-Si investigated in the present work, the impact of these impurities is smaller due to the use of the high-purity crystallization crucible. C. Comparison with standard mc silicon Recently we have shown that standard mc n-type silicon features significantly higher material quality and efficiency potential compared with standard mc p-type silicon of comparable impurity content crystallized under identical conditions [8]. On the first glance, the results discussed in the previous sections seem to contradict those findings. However, it has to be considered that beside the granular seed layer assisting the growth of the HPM-Si blocks investigated in this work, also different crucibles were used for crystallization. The standard mc silicon blocks investigated in [8] were crystallized in G2 size crucibles of standard industrial quality, whereas the HPM-Si blocks were grown in G1 size crucibles of high purity. Thus, metal impurity concentrations in the HPM-Si blocks are lower than in the standard mc Si blocks of reference [8]. The wafers investigated in [8] additionally featured an edge region of low material quality affected by solid-state in-diffusion of impurities from the crucible, where Fei is the dominant defect in the p-type wafers and FeSi2precipitates in the n-type wafers [26]. In contrast, due to the crystallization in high-purity crucibles, the HPM-Si blocks feature a significantly smaller edge region, which has been cut off entirely and, consequently, does not affect the material quality of the wafers investigated here (cf. Fig 1). Considering these differences, we can conclude that for mc silicon with a higher concentration of dissolved metal impurities, such as edge wafers or wafers from blocks crystallized in crucibles of standard quality in industrial production, a gain in material quality and efficiency potential could be achieved by n-type doping. Of course, size scaling effects for larger industrial crystallizations have to be taken into account [11]. In this work, the higher material quality is notable only for processes without P-diffusion, as dissolved impurities become less important after P-gettering, thus leading to comparable diffusion lengths in the investigated p- and n-type HPM-Si after processes including a phosphorus diffusion. IV. HIGH EFFICIENCY MC SILICON SOLAR CELLS A. Applicability of TOPCon to mc silicon substrate For the fabrication of high-efficiency mc silicon solar cells, a high-efficiency cell concept is needed, which is able to exploit the high electrical material quality of HPM-Si. A promising cell concept with a full-area Tunnel Oxide Passivated rear Contact (TOPCon) [13] has been developed at Fraunhofer ISE in the past years, featuring high efficiencies of 6 up to 24.9% on monocrystalline n-type silicon [28]. However, the applicability of this cell concept to multicrystalline silicon has not been proven so far, and depends crucially on the following issues: Does the TOPCon-layer passivate the mc silicon surface? The passivation quality might depend on the different crystal orientation of the grains due to an inhomogeneous thickness of the tunnel oxide. Is the contact resistance of the TOPCon-layer on the mc silicon surface low enough? The tunnel transport might be affected by different thicknesses of the tunnel oxide in different grains. Is the material quality of the mc silicon bulk after the high-temperature steps of the TOPCon process high enough for the fabrication of high-efficiency solar cells? In order to answer the first two questions, symmetric lifetime samples were fabricated, passivated with a TOPConlayer on both chemically polished sides of the wafer. After deposition of the layer (ultra-thin wet chemical oxide layer (~14 Å) in combination with a 15 nm thick phosphorus-doped Si layer deposited on top of the oxide), the samples were subjected to an annealing step at 8 C for one hour. Calibrated PL imaging at an illumination of one sun reveals homogeneous minority charge carrier diffusion lengths within the grains on a high level of up to 8 µm (cf. Fig. 5), which proves the good passivation quality of the TOPCon-layer on the chemically polished mc-si surface, irrespective of the crystal orientation. Note that the thickness was adapted to monocrystalline <1> silicon and that a possible thickness variation of the layer depending on grain orientation was not investigated. Minority charge carrier diffusion length Ldiff (µm) Fig. 5. Diffusion length image (at 1 sun illumination) of an n-type HPM-Si sample symmetrically passivated with a TOPCon-layer, revealing a good passivation quality of the layer on the chemically polished mc silicon surface. Measured contact resistivities on TLM-structures (transmission line model measurements, [29]) on the same samples (2-3 mωcm²) are even lower than results obtained on FZ reference samples (~1 mωcm²), which indicates that no problems related to high contact resistances should occur on mc silicon surfaces. The third question regarding the material quality can partly be answered by the results of section III. Fig. 3 shows that average minority carrier diffusion lengths in n-type HPM-Si

7 (c) 215 IEEE. DOI: 1.119/JPHOTOV Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material 7 are on a very high level after a TOPCon sequence including a P-gettering step (> 6 µm). Even after a boron diffusion without a further P-diffusion gettering step, average diffusion lengths exceed 5 µm. The additional annealing step at 8 C, which is included in the standard TOPCon process sequence (cf. Fig. 2b), does not change the material quality compared to the single boron diffusion (results not shown in Fig. 3). Thus, the material quality of the investigated n-type HPM-Si should be high enough for the fabrication of highefficiency solar cells. The results on passivation quality, contact resistance, and material quality show that there are no principal restrictions for the transfer of the TOPCon concept to n-type HPM-Si. In order to assess the material-related efficiency losses quantitatively, an ELBA analysis is performed in the next section, before the first cell results are presented in section IV.C. B. Efficiency potential As explained in section II.B, predictions of the efficiency potential are performed by an ELBA analysis for three cases: i) Standard TOPCon process (+) on isotextured surface (cell limit: 2.9%) ii) TOPCon process including additional P-gettering (B+P+) on isotextured surface (cell limit: 2.9%) iii) TOPCon process including additional P-gettering (B+P+) with optimized surface texture (cell limit: 22.4%) The first analysis delivers a realistic estimation of the efficiency potential after the standard TOPCon process, whereas the second analysis gives access to the efficiency gain due to the better material quality after P-gettering. Finally, the third analysis is included to obtain an estimation of the efficiency gain by improvements of the front surface reflectance, e.g. by applying a honeycomb texture. The better reflectance of the front surface alone increases the cell limit from 2.9% (case i and ii) to 22.4%. All analyses are performed on large-area wafers (round 4-inch wafers for case i, cm² wafers in ii and iii). The results of the ELBA analyses are summarized in Tab. I. For case i, the spatially resolved efficiency for the central 2 2 cm² area is shown in Fig. 6a, efficiency images for cases ii and iii can be found in a separate publication [27]. TABLE I Globally averaged cell parameters obtained from ELBA analyses i) ii) iii) V oc (mv) J sc (ma/cm²) FF bulk (%) η bulk (%) cell limit (%) η loss (% abs ) Applying the standard TOPCon process to the investigated n-type HPM-Si results in an efficiency potential 19.9%, which corresponds to a material-related loss of -1.% abs compared with the cell limit. The spatially resolved analysis allows for separately assessing losses due to homogeneously distributed recombination centers by evaluating the efficiency potential in grains, and losses due to decorated structural crystal defects by comparing the total losses with the efficiency losses in grains. Thus, after the standard TOPCon process, the major part of material-related efficiency losses can be attributed to decorated structural crystal defects (-.7% abs ), whereas losses due to homogeneously distributed recombination centers are low (-.3% abs ), leading to an efficiency potential close to the cell limit in the best grains. This is also highlighted in Fig. 6a. It should be noted that also the losses due to structural crystal defects are significantly lower than in standard mc n-type silicon [8], reflecting the good crystal quality of the investigated HPM-Si. Including an additional P-diffusion in the TOPCon process significantly reduces the losses due to decorated structural crystal defects to -.5% abs and losses due to homogeneously distributed recombination centers to a very low level (-.2% abs ), which results in a larger efficiency potential of 2.2%. Finally, an improvement of the front surface reflectance leads to a strong increase in J sc, while the materialrelated losses are unchanged. Thus, the increase in the cell limit (+1.5% abs ) would lead to an efficiency potential of 21.7% for n-type HPM-Si TOPCon solar cells featuring a honeycomb texture on the front side. Both results support the high efficiency potential of n-type HPM-Si with material-related losses of only.7% abs. C. Solar cell characterization The promising results of the pre-characterization of section IV.A and the predicted efficiency potential of section IV.B encouraged us to fabricate TOPCon solar cells according to the standard processing of Fig. 2b on n-type HPM-Si substrate. From the ELBA simulations, we would expect an efficiency potential of 19.9% for this process on large area 4- inch wafers. Seven 2 2 cm² solar cells were fabricated per wafer as described in section II.A (cf. Fig. 1). Thus, the efficiency potential on the cell areas could slightly differ from the average efficiency potential of the whole wafer. However, due to the very homogeneous material quality (cf. Fig. 5), these deviations are expected to be small, such that the efficiency of 2 2 cm² solar cells is representative for the efficiency of solar cells on large area. This is also confirmed by an ELBA analysis of a 2 2 cm² area (corresponding to the region of the best solar cell), which predicts an efficiency potential of 2.% after the standard TOPCon process. The predicted solar cell parameters, which only slightly differ from the predictions on large area (cf. Tab. I), are included in Tab. II, and the spatially resolved efficiency prediction is shown in Fig. 6a. The best multicrystalline n-type TOPCon solar cell after the standard TOPCon process features an efficiency of 19.3%, which was independently confirmed by Fraunhofer ISE Callab (cf. Tab. II, ARC). The measured V oc of 663. mv is in excellent agreement with the value predicted by ELBA, and in

8 (c) 215 IEEE. DOI: 1.119/JPHOTOV Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material TABLE II Result of an ELBA analysis on a representative 2 2 cm² area, and solar cell results of the best multicrystalline n-type TOPCon solar cell featuring a standard anti-reflection coating (ARC, independently confirmed) and a double anti-reflection coating (DARC) ELBA Best solar Best solar prediction cell (ARC) cell (DARC) Voc (mv) Jsc (ma/cm²) FF (%) η (%) Predicted Efficiency Potential TOPCon isotextured ηaverage = 2.% η (%) 2.9 Jsc Map Best solar cell (DARC) Jsc, avg. = 38.7 ma/cm² a) ma cm² cm Jsc 36 b) 34.9 Fig. 6. a) ELBA prediction of the spatially resolved efficiency potential after the standard TOPCon process for the 2 2 cm² region corresponding to the cell area. b) Spatially resolved Jsc of the best solar cell with DARC, obtained from Jsc-density mapping [3]. Slight differences in crystal structure visible in the two images are due to a small difference in block position. the range of world record mc silicon solar cells. For comparison, the world record mc-si solar cell presented by TRINA solar (η = 2.8%) [1] features a similar Voc of mv [9], as well as the previous world record mc-si solar cell from Schultz et al. [31] (η = 2.4%; Voc = 664 mv) [32]. Also, the measured Jsc of 38.4 ma/cm² is on a high level and in reasonable agreement with the ELBA prediction. However, our solar cell suffers from a rather low fill factor of only 76.4%, which is significantly lower than the predicted value of 8.7%. This results in the measured efficiency of 19.3% instead of the predicted efficiency of 2.%. We can attribute the low fill factor to high series resistance losses in the front metal grid due to processing issues during plating, which we also observed on FZ reference cells. Additionally, measured pseudo-fill factors on mc n-type TOPCon solar cells were up to 81.8%. From these results, a material-related issue can be excluded, and we expect even higher solar cell efficiencies in the range of 2% for the same process and material without problems during plating. Additionally, as discussed before, including a P-gettering diffusion in the TOPCon solar cell process would decrease the material-related losses by ~.3%abs (cf. Tab. I). Furthermore, a large increase in Jsc could be achieved by an advanced front surface texture instead of the applied acidic texture, which would significantly increase the solar cell efficiency by approximately 1.5%abs. Thus, accounting for these two aspects, efficiencies of 21-22% should be attainable on n-type HPM-Si TOPCon solar cells (cf. Tab. I), which is a very promising prospect for 8 multicrystalline n-type silicon solar cells. A slight improvement of the front surface reflectance can already be achieved by a double anti-reflection coating, which we applied to the front side of the record mc n-type TOPCon solar cell in a second step. This leads to an increase in Jsc to a value of ma/cm² and, thus, to an efficiency of 19.6% for the same cell (cf. Tab. II, DARC). To our knowledge, this is the highest efficiency reported for a solar cell on multicrystalline n-type silicon substrate, exceeding previously published values by more than 3%abs [33, 34]. In order to understand the losses occurring in this solar cell compared with the Auger lifetime limit, we performed a complete cell analysis as described in detail in [35]. Good agreement between global cell results (cf. Tab. II, DARC) and average values for Jsc (38.7 ma/cm²), Voc (663.9 mv), and fill factor (75.3%) obtained from the different imaging methods included in this loss analysis is observed. This confirms the applicability of such analysis for a spatially resolved insight into the specific loss mechanisms. Exemplarily, Fig. 6b shows the Jsc map of the solar cell measured by short-circuit current density mapping [3], which nicely reveals the Jsc-losses at grain boundaries. Small differences in the crystal structure visible in the Jsc-map of the solar cell and the predicted efficiency image from Fig. 6a are due to slightly different block heights of the lifetime sample and the solar cell. In total, compared with the Auger-limit, 15% of the efficiency losses can be attributed to series resistance losses and 2% to optical losses, whereas the major part is due to recombination losses in the emitter and the bulk. As outlined before, the series resistance losses should already be significantly reduced without processing issues during plating, and the application of an improved front surface texture, e.g. a honeycomb texture, would strongly reduce the optical losses. These will be the next steps towards highly efficient multicrystalline n-type silicon solar cells beyond 21%. V. CONCLUSION In this work, we presented a thorough study on material quality and efficiency potential of n-type high performance multicrystalline silicon. By measuring the concentrations of interstitial iron and chromium at a p-type sister block, we can attribute the higher minority carrier diffusion length in asgrown n-type HPM-Si to its smaller sensitivity to interstitial iron. This is also the reason for the significantly larger minority charge carrier diffusion length in n-type HPM-Si after the investigated boron diffusion. In the initial state, the relative impact of chromium point defects on recombination is comparable in both p- and n-type HPM-Si. After P-diffusion gettering, both materials feature diffusion lengths on a similarly high level. The low impact of dissolved impurities after gettering is partly attributed to the use of high-purity crucibles for crystallization. The applicability of the high-efficiency TOPCon cell concept to multicrystalline silicon was demonstrated, and a high efficiency potential of n-type HPM-Si after the TOPCon cell process was confirmed by an ELBA analysis, predicting efficiencies exceeding 2% on isotextured samples and close

9 (c) 215 IEEE. DOI: 1.119/JPHOTOV Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material 9 to 22% on samples with an improved texture. Finally, we presented a multicrystalline n-type TOPCon solar cell featuring an efficiency of 19.6%, which is a record efficiency reported for a solar cell on multicrystalline n-type silicon substrate. By improving the optics and avoiding large series resistance losses, efficiencies exceeding 21% should be attainable on mc n-type TOPCon solar cells based on the material investigated here. ACKNOWLEDGMENT The authors would like to thank A. Leimenstoll, F. Schätzle, S. Seitz, and N. Weber for sample preparation as well as E. Schäffer for measuring the solar cells. REFERENCES [1] D. Macdonald and L. J. Geerligs, "Recombination activity of interstitial iron and other transition metal point defects in p- and n-type crystalline silicon," Appl. Phys. Lett., vol. 85, pp , 24. [2] J. Schmidt, A. G. Aberle, and R. Hezel, "Investigation of carrier lifetime instabilities in Cz-grown silicon," in Proc. 26th IEEE Photovoltaic Spec. Conf., Anaheim, California, USA, 1997, pp [3] M. Acciarri, S. Binetti, A. Le Donne, S. Marchionna, M. Vimercati, J. Libal, R. Kopecek, and K. Wambach, "Effect of P-induced gettering on extended defects in n-type multicrystalline silicon," Prog. Photovoltaics, Res. Appl., vol. 15, pp , 27. [4] L. J. Geerligs, Y. Komatsu, I. Röver, K. Wambach, I. Yamaga, and T. Saitoh, "Precipitates and hydrogen passivation at crystal defects in n- and p-type multicrystalline silicon," J. Appl. Phys., vol. 12, p. 9372, 27. [5] W. Kwapil, J. Schön, F. Schindler, W. Warta, and M. C. Schubert, "Impact of Iron Precipitates on Carrier Lifetime in As-Grown and Phosphorus-Gettered Multicrystalline Silicon Wafers in Model and Experiment," IEEE J. Photovoltaics, vol. 4, pp , 214. [6] W. Kwapil, J. Schön, W. Warta, and M. C. Schubert, "Carrier Recombination at Metallic Precipitates in p- and n-type Silicon," IEEE J. Photovoltaics, vol. 5, pp , 215. [7] A. Cuevas, M. J. Kerr, C. Samundsett, F. Ferrazza, and G. Coletti, "Millisecond minority carrier lifetimes in n-type multicrystalline silicon," Appl. Phys. Lett., vol. 81, pp , 22. [8] F. Schindler, B. Michl, A. Kleiber, H. Steinkemper, J. Schon, W. Kwapil, P. Krenckel, S. Riepe, W. Warta, and M. C. Schubert, "Potential Gain in Multicrystalline Silicon Solar Cell Efficiency by n-type Doping," IEEE J. Photovoltaics, vol. 5, pp , 215. [9] M. A. Green, K. Emery, Y. Hishikawa, W. Warta, and E. D. Dunlop, "Solar cell efficiency tables (Version 45)," Prog. Photovoltaics, Res. Appl., vol. 23, pp. 1-9, 215. [1] P. Verlinden, W. Deng, X. Zhang, Y. Yang, J. Xu, Y. Shu, P. Quan, J. Sheng, S. Zhang, J. Bao, F. Ping, Y. Zhang, and Z. Feng, "Strategy, Development and Mass Production of High-Efficiency Crystalline Si PV Modules," presented at the 6th World Conf. PV Energy Convers., Kyoto, Japan, 214. [11] M. C. Schubert, J. Schön, F. Schindler, W. Kwapil, A. Abdollahinia, B. Michl, S. Riepe, C. Schmid, M. Schumann, S. Meyer, and W. Warta, "Impact of Impurities From Crucible and Coating on mc-silicon Quality - the Example of Iron and Cobalt," IEEE J. Photovoltaics, vol. 3, pp , 213. [12] F. Schindler, B. Michl, J. Schön, W. Kwapil, W. Warta, and M. C. Schubert, "Solar cell efficiency losses due to impurities from the crucible in multicrystalline silicon," IEEE J. Photovoltaics, vol. 4, pp , 214. [13] F. Feldmann, M. Bivour, C. Reichel, H. Steinkemper, M. Hermle, and S. W. Glunz, "Tunnel oxide passivated contacts as an alternative to partial rear contacts," Sol. Energy Mater. Sol. Cells, vol. 131, pp. 46-5, 214. [14] B. Michl, M. Rüdiger, J. Giesecke, M. Hermle, W. Warta, and M. C. Schubert, "Efficiency limiting bulk recombination in multicrystalline silicon solar cells," Sol. Energy Mater. Sol. Cells, vol. 98, pp , 212. [15] F. Feldmann, M. Bivour, C. Reichel, M. Hermle, and G. S. W., "A passivated rear contact for high-efficiency n-type silicon solar cells enabling high V ocs and FF>82%," in Proc. 28th Eur. Photovoltaic Sol. Energy Conf. Exhib., Paris, France, 213, pp [16] J. A. Giesecke, M. C. Schubert, B. Michl, F. Schindler, and W. Warta, "Minority carrier lifetime imaging of silicon wafers calibrated by quasisteady-state photoluminescence," Sol. Energy Mater. Sol. Cells, vol. 95, pp , 211. [17] D. Macdonald, J. Tan, and T. Trupke, "Imaging interstitial iron concentrations in boron-doped crystalline silicon unsing photoluminescence," J. Appl. Phys., vol. 13, p. 7371, 28. [18] M. C. Schubert, H. Habenicht, and W. Warta, "Imaging of metastable defects in silicon," IEEE J. Photovoltaics, vol. 1, pp , 211. [19] A. A. Istratov, H. Hieslmair, and E. R. Weber, "Iron and its complexes in silicon," Appl. Phys. A (Mater. Sci. Process.), vol. A69, pp , [2] D. Macdonald, T. Roth, P. N. K. Deenapanray, T. Trupke, and R. A. Bardos, "Doping dependence of the carrier lifetime crossover point upon dissociation of iron-boron pairs in cystalline silicon," Appl. Phys. Lett., vol. 89, pp , 26. [21] C. Sun, F. E. Rougieux, and D. Macdonald, "Reassessment of the recombination parameters of chromium in n- and p-type crystalline silicon and chromium-boron pairs in p-type crystalline silicon," J. Appl. Phys., vol. 115, p , 214. [22] B. Michl, M. Kasemann, W. Warta, and M. C. Schubert, "Wafer thickness optimization for silicon solar cells of heterogeneous material quality," Phys. Status Solidi RRL Rapid Res. Lett., vol. 7, pp , 213. [23] S. P. Phang and D. Macdonald, "Direct comparison of boron, phosphorus, and aluminum gettering of iron in crystalline silicon," J. Appl. Phys., vol. 19, p , 211. [24] J. Schön, H. Habenicht, W. Warta, and M. C. Schubert, "Chromium distribution in multicrystalline silicon: comparison of simulations and experiments," Prog. Photovoltaics, Res. Appl., vol. 21, pp , 213. [25] J. Schmidt, B. Lim, D. Walter, K. Bothe, S. Gatz, T. Dullweber, and P. P. Altermatt, "Impurity-related limitations of next-generation industrial silicon solar cells," IEEE J. Photovoltaics, vol. 3, pp , 213. [26] J. Schön, F. Schindler, W. Kwapil, M. Knörlein, P. Krenckel, S. Riepe, W. Warta, and M. C. Schubert, "Identification of the most relevant metal impurities in mc n-type silicon for solar cells," Sol. Energy Mater. Sol. Cells, vol. 142, pp , 215. [27] F. Schindler, B. Michl, P. Krenckel, S. Riepe, F. Feldmann, J. Benick, W. Warta, and M. C. Schubert, "Efficiency Potential of p- and n-type High Performance Multicrystalline Silicon," Energy Procedia, vol. 77, pp , 215. [28] M. Hermle, F. Feldmann, J. Eisenlohr, J. Benick, A. Richter, B. Lee, P. Stradins, A. Rohatgi, and S. W. Glunz, "Approaching Efficiencies above 25% with both Sides-contacted Silicon Solar Cells," IEEE J. Photovoltaics, submitted for publication, 215. [29] G. K. Reeves and H. B. Harrison, "Obtaining the specific contact resistance from transmission line model measurements," IEEE Electron Device Lett., vol. 3, pp , [3] M. Padilla, B. Michl, B. Thaidigsmann, W. Warta, and M. C. Schubert, "Short-circuit current density mapping for solar cells," Sol. Energy Mater. Sol. Cells, vol. 12, Part A, pp , 214. [31] O. Schultz, S. W. Glunz, and G. Willeke, "Multicrystalline silicon solar cells exceeding 2% efficiency," Prog. Photovoltaics, Res. Appl., vol. 12, pp , 24. [32] M. A. Green, K. Emery, Y. Hishikawa, W. Warta, and E. D. Dunlop, "Solar cell efficiency tables (version 44)," Prog. Photovoltaics, Res. Appl., vol. 22, pp , 214. [33] V. D. Mihailetchi, L. J. Geerligs, Y. Komatsu, T. Buck, I. Röver, K. Wambach, C. Knopf, and R. Kopecek, "High Efficiency Industrial Screen Printed n-type mc-si Solar Cells With Front Boron Emitter," in Proc. 22nd Eur. Photovoltaic Sol. Energy Conf., Milan, Italy, 27, pp [34] Y. Komatsu, V. D. Mihailetchi, L. J. Geerligs, B. van Dijk, J. B. Rem, and M. Harris, "Homogeneous p+ emitter diffused using boron tribromide for record 16.4% screen-printed large area n-type mc-si solar cell," Sol. Energy Mater. Sol. Cells, vol. 93, pp , 29. [35] B. Michl, M. Padilla, I. Geisemeyer, S. T. Haag, F. Schindler, M. C. Schubert, and W. Warta, "Imaging Techniques for Quantitative Silicon Material and Solar Cell Analysis," IEEE J. Photovoltaics, vol. 4, pp , 214.